fix(task_05_fmops): upgrade judge model from legacy Sonnet 4 to Sonnet 4.6

[el-pedrito]

Problem

The QualitativeModelEvaluation pipeline step fails with:

Float cannot represent non numeric value: nan

This error occurs 6 times (3 metrics × 2 aggregations: mean + variance).

Root Cause

The LLM-as-a-judge model claude-sonnet-4-20250514-v1:0 has been marked as Legacy by Anthropic on Amazon Bedrock. When a model is Legacy and hasn't been used in the last 30 days, Bedrock rejects all invocations with:

ResourceNotFoundException: Access denied. This Model is marked by provider as Legacy
and you have not been actively using the model in the last 30 days.

This causes make_genai_metric to silently fail on every judge call → all scores become None → aggregated as NaN → the SageMaker Managed MLflow tracking server (GraphQL API) rejects NaN float values.

The downstream EvaluationGate step also fails with Incompatible types in BinaryCondition left type [String], right type [Float] because the qualitative result is not a valid float.

Fix

Replace all references to the legacy model ID with the active claude-sonnet-4-6:

• bedrock:/global.anthropic.claude-sonnet-4-20250514-v1:0 → bedrock:/global.anthropic.claude-sonnet-4-6

Also removed commented-out references to the legacy claude-3-haiku-20240307-v1:0.

Files Changed

• steps/qualitative_eval_step.py — 3 metric definitions + 1 log_param
• 05.00_fmops_examples.ipynb — judge_llm variable + markdown description

Testing

Pipeline executed end-to-end successfully after the fix:

• QualitativeModelEvaluation: ✅ Succeeded (real scores logged)
• EvaluationGate: ✅ Succeeded
• ModelRegistration: ✅ Succeeded